Reinforcement learning-based mapless navigation holds significant potential. However, it faces challenges in indoor environments with local minima area. This paper introduces a safe mapless navigation framework utilizing hierarchical reinforcement learning (HRL) to enhance navigation through such areas. The high-level policy creates a sub-goal to direct the navigation process. Notably, we have developed a sub-goal update mechanism that considers environment congestion, efficiently avoiding the entrapment of the robot in local minimum areas. The low-level motion planning policy, trained through safe reinforcement learning, outputs real-time control instructions based on acquired sub-goal. Specifically, to enhance the robot's environmental perception, we introduce a new obstacle encoding method that evaluates the impact of obstacles on the robot's motion planning. To validate the performance of our HRL-based navigation framework, we conduct simulations in office, home, and restaurant environments. The findings demonstrate that our HRL-based navigation framework excels in both static and dynamic scenarios. Finally, we implement the HRL-based navigation framework on a TurtleBot3 robot for physical validation experiments, which exhibits its strong generalization capabilities.
Control Lyapunov functions are traditionally used to design a controller which ensures convergence to a desired state, yet deriving these functions for nonlinear systems remains a complex challenge. This paper presents a novel, sample-efficient method for neural approximation of nonlinear Lyapunov functions, leveraging self-supervised Reinforcement Learning (RL) to enhance training data generation, particularly for inaccurately represented regions of the state space. The proposed approach employs a data-driven World Model to train Lyapunov functions from off-policy trajectories. The method is validated on both standard and goal-conditioned robotic tasks, demonstrating faster convergence and higher approximation accuracy compared to the state-of-the-art neural Lyapunov approximation baseline. The code is available at: https://github.com/CAV-Research-Lab/SACLA.git
The linear regression (LR) method offers the advantage that optimal parameters can be calculated relatively easily, although its representation capability is limited than that of the deep learning technique. To improve deep reinforcement learning, the Least Squares Deep Q Network (LS-DQN) method was proposed by Levine et al., which combines Deep Q Network (DQN) with LR method. However, the LS-DQN method assumes that the actions are discrete. In this study, we propose the Double Least Squares Deep Deterministic Policy Gradient (DLS-DDPG) method to address this limitation. This method combines the LR method with the Deep Deterministic Policy Gradient (DDPG) technique, one of the representative deep reinforcement learning algorithms for continuous action cases. Numerical experiments conducted in MuJoCo environments showed that the LR update improved performance at least in some tasks, although there are difficulties such as the inability to make the regularization terms small.
Humanoid robots are anticipated to acquire a wide range of locomotion capabilities while ensuring natural movement across varying speeds and terrains. Existing methods encounter a fundamental dilemma in learning humanoid locomotion: reinforcement learning with handcrafted rewards can achieve agile locomotion but produces unnatural gaits, while Generative Adversarial Imitation Learning (GAIL) with motion capture data yields natural movements but suffers from unstable training processes and restricted agility. Integrating these approaches proves challenging due to the inherent heterogeneity between expert policies and human motion datasets. To address this, we introduce StyleLoco, a novel two-stage framework that bridges this gap through a Generative Adversarial Distillation (GAD) process. Our framework begins by training a teacher policy using reinforcement learning to achieve agile and dynamic locomotion. It then employs a multi-discriminator architecture, where distinct discriminators concurrently extract skills from both the teacher policy and motion capture data. This approach effectively combines the agility of reinforcement learning with the natural fluidity of human-like movements while mitigating the instability issues commonly associated with adversarial training. Through extensive simulation and real-world experiments, we demonstrate that StyleLoco enables humanoid robots to perform diverse locomotion tasks with the precision of expertly trained policies and the natural aesthetics of human motion, successfully transferring styles across different movement types while maintaining stable locomotion across a broad spectrum of command inputs.
Inference scaling empowers LLMs with unprecedented reasoning ability, with reinforcement learning as the core technique to elicit complex reasoning. However, key technical details of state-of-the-art reasoning LLMs are concealed (such as in OpenAI o1 blog and DeepSeek R1 technical report), thus the community still struggles to reproduce their RL training results. We propose the $\textbf{D}$ecoupled Clip and $\textbf{D}$ynamic s$\textbf{A}$mpling $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{DAPO}$) algorithm, and fully open-source a state-of-the-art large-scale RL system that achieves 50 points on AIME 2024 using Qwen2.5-32B base model. Unlike previous works that withhold training details, we introduce four key techniques of our algorithm that make large-scale LLM RL a success. In addition, we open-source our training code, which is built on the verl framework, along with a carefully curated and processed dataset. These components of our open-source system enhance reproducibility and support future research in large-scale LLM RL.
Reinforcement learning control algorithms face significant challenges due to out-of-distribution and inefficient exploration problems. While model-based reinforcement learning enhances the agent's reasoning and planning capabilities by constructing virtual environments, training such virtual environments can be very complex. In order to build an efficient inference model and enhance the representativeness of learning data, we propose the Counterfactual Experience Augmentation (CEA) algorithm. CEA leverages variational autoencoders to model the dynamic patterns of state transitions and introduces randomness to model non-stationarity. This approach focuses on expanding the learning data in the experience pool through counterfactual inference and performs exceptionally well in environments that follow the bisimulation assumption. Environments with bisimulation properties are usually represented by discrete observation and action spaces, we propose a sampling method based on maximum kernel density estimation entropy to extend CEA to various environments. By providing reward signals for counterfactual state transitions based on real information, CEA constructs a complete counterfactual experience to alleviate the out-of-distribution problem of the learning data, and outperforms general SOTA algorithms in environments with difference properties. Finally, we discuss the similarities, differences and properties of generated counterfactual experiences and real experiences. The code is available at https://github.com/Aegis1863/CEA.
Recent research has highlighted the significance of natural language in enhancing the controllability of generative models. While various efforts have been made to leverage natural language for content generation, research on deep reinforcement learning (DRL) agents utilizing text-based instructions for procedural content generation remains limited. In this paper, we propose IPCGRL, an instruction-based procedural content generation method via reinforcement learning, which incorporates a sentence embedding model. IPCGRL fine-tunes task-specific embedding representations to effectively compress game-level conditions. We evaluate IPCGRL in a two-dimensional level generation task and compare its performance with a general-purpose embedding method. The results indicate that IPCGRL achieves up to a 21.4% improvement in controllability and a 17.2% improvement in generalizability for unseen instructions. Furthermore, the proposed method extends the modality of conditional input, enabling a more flexible and expressive interaction framework for procedural content generation.
Reinforcement learning (RL) has increasingly been applied to solve real-world planning problems, with progress in handling large state spaces and time horizons. However, a key bottleneck in many domains is that RL methods cannot accommodate large, combinatorially structured action spaces. In such settings, even representing the set of feasible actions at a single step may require a complex discrete optimization formulation. We leverage recent advances in embedding trained neural networks into optimization problems to propose SEQUOIA, an RL algorithm that directly optimizes for long-term reward over the feasible action space. Our approach embeds a Q-network into a mixed-integer program to select a combinatorial action in each timestep. Here, we focus on planning over restless bandits, a class of planning problems which capture many real-world examples of sequential decision making. We introduce coRMAB, a broader class of restless bandits with combinatorial actions that cannot be decoupled across the arms of the restless bandit, requiring direct solving over the joint, exponentially large action space. We empirically validate SEQUOIA on four novel restless bandit problems with combinatorial constraints: multiple interventions, path constraints, bipartite matching, and capacity constraints. Our approach significantly outperforms existing methods -- which cannot address sequential planning and combinatorial selection simultaneously -- by an average of 24.8\% on these difficult instances.
One of the challenges of aligning large models with human preferences lies in both the data requirements and the technical complexities of current approaches. Predominant methods, such as RLHF, involve multiple steps, each demanding distinct types of data, including demonstration data and preference data. In RLHF, human preferences are typically modeled through a reward model, which serves as a proxy to guide policy learning during the reinforcement learning stage, ultimately producing a policy aligned with human preferences. However, in this paper, we propose a fresh perspective on learning alignment based on inverse reinforcement learning principles, where the optimal policy is still derived from reward maximization. However, instead of relying on preference data, we directly learn the reward model from demonstration data. This new formulation offers the flexibility to be applied even when only demonstration data is available, a capability that current RLHF methods lack, and it also shows that demonstration data offers more utility than what conventional wisdom suggests. Our extensive evaluation, based on public reward benchmark, HuggingFace Open LLM Leaderboard and MT-Bench, demonstrates that our approach compares favorably to state-of-the-art methods that rely solely on demonstration data.
We study offline reinforcement learning problems with a long-run average reward objective. The state-action pairs generated by any fixed behavioral policy thus follow a Markov chain, and the {\em empirical} state-action-next-state distribution satisfies a large deviations principle. We use the rate function of this large deviations principle to construct an uncertainty set for the unknown {\em true} state-action-next-state distribution. We also construct a distribution shift transformation that maps any distribution in this uncertainty set to a state-action-next-state distribution of the Markov chain generated by a fixed evaluation policy, which may differ from the unknown behavioral policy. We prove that the worst-case average reward of the evaluation policy with respect to all distributions in the shifted uncertainty set provides, in a rigorous statistical sense, the least conservative estimator for the average reward under the unknown true distribution. This guarantee is available even if one has only access to one single trajectory of serially correlated state-action pairs. The emerging robust optimization problem can be viewed as a robust Markov decision process with a non-rectangular uncertainty set. We adapt an efficient policy gradient algorithm to solve this problem. Numerical experiments show that our methods compare favorably against state-of-the-art methods.
We introduce a Reinforcement Learning (RL)-based method for re-synthesis of quantum circuits containing arbitrary Pauli rotations alongside Clifford operations. By collapsing each sub-block to a compact representation and then synthesizing it step-by-step through a learned heuristic, we obtain circuits that are both shorter and compliant with hardware connectivity constraints. We find that the method is fast enough and good enough to work as an optimization procedure: in direct comparisons on 6-qubit random Pauli Networks against state-of-the-art heuristic methods, our RL approach yields over 2x reduction in two-qubit gate count, while executing in under 10 milliseconds per circuit. We further integrate the method into a collect-and-re-synthesize pipeline, applied as a Qiskit transpiler pass, where we observe average improvements of 20% in two-qubit gate count and depth, reaching up to 60% for many instances, across the Benchpress benchmark. These results highlight the potential of RL-driven synthesis to significantly improve circuit quality in realistic, large-scale quantum transpilation workloads.
Generating molecules with desired chemical properties presents a critical challenge in fields such as chemical synthesis and drug discovery. Recent advancements in artificial intelligence (AI) and deep learning have significantly contributed to data-driven molecular generation. However, challenges persist due to the inherent sensitivity of simplified molecular input line entry system (SMILES) representations and the difficulties in applying generative adversarial networks (GANs) to discrete data. This study introduces RL-MolGAN, a novel Transformer-based discrete GAN framework designed to address these challenges. Unlike traditional Transformer architectures, RL-MolGAN utilizes a first-decoder-then-encoder structure, facilitating the generation of drug-like molecules from both $de~novo$ and scaffold-based designs. In addition, RL-MolGAN integrates reinforcement learning (RL) and Monte Carlo tree search (MCTS) techniques to enhance the stability of GAN training and optimize the chemical properties of the generated molecules. To further improve the model's performance, RL-MolWGAN, an extension of RL-MolGAN, incorporates Wasserstein distance and mini-batch discrimination, which together enhance the stability of the GAN. Experimental results on two widely used molecular datasets, QM9 and ZINC, validate the effectiveness of our models in generating high-quality molecular structures with diverse and desirable chemical properties.
Lifelong Reinforcement Learning (LRL) holds significant potential for addressing sequential tasks, but it still faces considerable challenges. A key difficulty lies in effectively preventing catastrophic forgetting and facilitating knowledge transfer while maintaining reliable decision-making performance across subsequent tasks in dynamic environments. To tackle this, we propose a novel framework, SDW (Similarity-Driven Weighting Framework), which leverages large-language-model-generated dynamic functions to precisely control the training process. The core of SDW lies in two functions pre-generated by large models: the task similarity function and the weight computation function. The task similarity function extracts multidimensional features from task descriptions to quantify the similarities and differences between tasks in terms of states, actions, and rewards. The weight computation function dynamically generates critical training parameters based on the similarity information, including the proportion of old task data stored in the Replay Buffer and the strategy consistency weight in the loss function, enabling an adaptive balance between learning new tasks and transferring knowledge from previous tasks. By generating function code offline prior to training, rather than relying on large-model inference during the training process, the SDW framework reduces computational overhead while maintaining efficiency in sequential task scenarios. Experimental results on Atari and MiniHack sequential tasks demonstrate that SDW significantly outperforms existing lifelong reinforcement learning methods.
With the increasing demand for efficient and flexible robotic exploration solutions, Reinforcement Learning (RL) is becoming a promising approach in the field of autonomous robotic exploration. However, current RL-based exploration algorithms often face limited environmental reasoning capabilities, slow convergence rates, and substantial challenges in Sim-To-Real (S2R) transfer. To address these issues, we propose a Curriculum Learning-based Transformer Reinforcement Learning Algorithm (CTSAC) aimed at improving both exploration efficiency and transfer performance. To enhance the robot's reasoning ability, a Transformer is integrated into the perception network of the Soft Actor-Critic (SAC) framework, leveraging historical information to improve the farsightedness of the strategy. A periodic review-based curriculum learning is proposed, which enhances training efficiency while mitigating catastrophic forgetting during curriculum transitions. Training is conducted on the ROS-Gazebo continuous robotic simulation platform, with LiDAR clustering optimization to further reduce the S2R gap. Experimental results demonstrate the CTSAC algorithm outperforms the state-of-the-art non-learning and learning-based algorithms in terms of success rate and success rate-weighted exploration time. Moreover, real-world experiments validate the strong S2R transfer capabilities of CTSAC.
Mobile robot navigation in dynamic environments with pedestrian traffic is a key challenge in the development of autonomous mobile service robots. Recently, deep reinforcement learning-based methods have been actively studied and have outperformed traditional rule-based approaches owing to their optimization capabilities. Among these, methods that assume a continuous action space typically rely on a Gaussian distribution assumption, which limits the flexibility of generated actions. Meanwhile, the application of diffusion models to reinforcement learning has advanced, allowing for more flexible action distributions compared with Gaussian distribution-based approaches. In this study, we applied a diffusion-based reinforcement learning approach to social navigation and validated its effectiveness. Furthermore, by leveraging the characteristics of diffusion models, we propose an extension that enables post-training action smoothing and adaptation to static obstacle scenarios not considered during the training steps.
Training robotic manipulation policies traditionally requires numerous demonstrations and/or environmental rollouts. While recent Imitation Learning (IL) and Reinforcement Learning (RL) methods have reduced the number of required demonstrations, they still rely on expert knowledge to collect high-quality data, limiting scalability and accessibility. We propose Sketch-to-Skill, a novel framework that leverages human-drawn 2D sketch trajectories to bootstrap and guide RL for robotic manipulation. Our approach extends beyond previous sketch-based methods, which were primarily focused on imitation learning or policy conditioning, limited to specific trained tasks. Sketch-to-Skill employs a Sketch-to-3D Trajectory Generator that translates 2D sketches into 3D trajectories, which are then used to autonomously collect initial demonstrations. We utilize these sketch-generated demonstrations in two ways: to pre-train an initial policy through behavior cloning and to refine this policy through RL with guided exploration. Experimental results demonstrate that Sketch-to-Skill achieves ~96% of the performance of the baseline model that leverages teleoperated demonstration data, while exceeding the performance of a pure reinforcement learning policy by ~170%, only from sketch inputs. This makes robotic manipulation learning more accessible and potentially broadens its applications across various domains.
DeepSeek-V3 and DeepSeek-R1 are leading open-source Large Language Models (LLMs) for general-purpose tasks and reasoning, achieving performance comparable to state-of-the-art closed-source models from companies like OpenAI and Anthropic -- while requiring only a fraction of their training costs. Understanding the key innovative techniques behind DeepSeek's success is crucial for advancing LLM research. In this paper, we review the core techniques driving the remarkable effectiveness and efficiency of these models, including refinements to the transformer architecture, innovations such as Multi-Head Latent Attention and Mixture of Experts, Multi-Token Prediction, the co-design of algorithms, frameworks, and hardware, the Group Relative Policy Optimization algorithm, post-training with pure reinforcement learning and iterative training alternating between supervised fine-tuning and reinforcement learning. Additionally, we identify several open questions and highlight potential research opportunities in this rapidly advancing field.
Uncertainty quantification is a critical aspect of reinforcement learning and deep learning, with numerous applications ranging from efficient exploration and stable offline reinforcement learning to outlier detection in medical diagnostics. The scale of modern neural networks, however, complicates the use of many theoretically well-motivated approaches such as full Bayesian inference. Approximate methods like deep ensembles can provide reliable uncertainty estimates but still remain computationally expensive. In this work, we propose contextual similarity distillation, a novel approach that explicitly estimates the variance of an ensemble of deep neural networks with a single model, without ever learning or evaluating such an ensemble in the first place. Our method builds on the predictable learning dynamics of wide neural networks, governed by the neural tangent kernel, to derive an efficient approximation of the predictive variance of an infinite ensemble. Specifically, we reinterpret the computation of ensemble variance as a supervised regression problem with kernel similarities as regression targets. The resulting model can estimate predictive variance at inference time with a single forward pass, and can make use of unlabeled target-domain data or data augmentations to refine its uncertainty estimates. We empirically validate our method across a variety of out-of-distribution detection benchmarks and sparse-reward reinforcement learning environments. We find that our single-model method performs competitively and sometimes superior to ensemble-based baselines and serves as a reliable signal for efficient exploration. These results, we believe, position contextual similarity distillation as a principled and scalable alternative for uncertainty quantification in reinforcement learning and general deep learning.
Target search problems are central to a wide range of fields, from biological foraging to the optimization algorithms. Recently, the ability to reset the search has been shown to significantly improve the searcher's efficiency. However, the optimal resetting strategy depends on the specific properties of the search problem and can often be challenging to determine. In this work, we propose a reinforcement learning (RL)-based framework to train agents capable of optimizing their search efficiency in environments by learning how to reset. First, we validate the approach in a well-established benchmark: the Brownian search with resetting. There, RL agents consistently recover strategies closely resembling the sharp resetting distribution, known to be optimal in this scenario. We then extend the framework by allowing agents to control not only when to reset, but also their spatial dynamics through turning actions. In this more complex setting, the agents discover strategies that adapt both resetting and turning to the properties of the environment, outperforming the proposed benchmarks. These results demonstrate how reinforcement learning can serve both as an optimization tool and a mechanism for uncovering new, interpretable strategies in stochastic search processes with resetting.
Reinforcement Learning (RL) has proven largely effective in obtaining stable locomotion gaits for legged robots. However, designing control algorithms which can robustly navigate unseen environments with obstacles remains an ongoing problem within quadruped locomotion. To tackle this, it is convenient to solve navigation tasks by means of a hierarchical approach with a low-level locomotion policy and a high-level navigation policy. Crucially, the high-level policy needs to be robust to dynamic obstacles along the path of the agent. In this work, we propose a novel way to endow navigation policies with robustness by a training process that models obstacles as adversarial agents, following the adversarial RL paradigm. Importantly, to improve the reliability of the training process, we bound the rationality of the adversarial agent resorting to quantal response equilibria, and place a curriculum over its rationality. We called this method Hierarchical policies via Quantal response Adversarial Reinforcement Learning (Hi-QARL). We demonstrate the robustness of our method by benchmarking it in unseen randomized mazes with multiple obstacles. To prove its applicability in real scenarios, our method is applied on a Unitree GO1 robot in simulation.